[Tarantool-patches] [PATCH 7/8] raft: fix crash on election timeout decrease

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Sun Dec 13 20:15:29 MSK 2020


If election timeout was decreased during election to a new value
making the current election expired immediately, it could crash in
libev.

Because it would mean the remaining time until election end became
negative. The negative timeout was passed to libev without any
checks, and there is an assertion, that a timeout should always
be >= 0.

Part of #5303
---
 src/lib/raft/raft.c   |  2 ++
 test/unit/raft.c      | 20 +++++++++++++++++++-
 test/unit/raft.result |  8 +++++---
 3 files changed, 26 insertions(+), 4 deletions(-)

diff --git a/src/lib/raft/raft.c b/src/lib/raft/raft.c
index ab007a462..4f6a5ee5e 100644
--- a/src/lib/raft/raft.c
+++ b/src/lib/raft/raft.c
@@ -895,6 +895,8 @@ raft_cfg_election_timeout(struct raft *raft, double timeout)
 		struct ev_loop *loop = raft_loop();
 		double timeout = raft_ev_timer_remaining(loop, &raft->timer) -
 				 raft->timer.at + raft->election_timeout;
+		if (timeout < 0)
+			timeout = 0;
 		raft_ev_timer_stop(loop, &raft->timer);
 		raft_ev_timer_set(&raft->timer, timeout, timeout);
 		raft_ev_timer_start(loop, &raft->timer);
diff --git a/test/unit/raft.c b/test/unit/raft.c
index b97d9d0aa..2c3935cbf 100644
--- a/test/unit/raft.c
+++ b/test/unit/raft.c
@@ -793,7 +793,7 @@ raft_test_heartbeat(void)
 static void
 raft_test_election_timeout(void)
 {
-	raft_start_test(11);
+	raft_start_test(13);
 	struct raft_node node;
 	raft_node_create(&node);
 
@@ -865,6 +865,24 @@ raft_test_election_timeout(void)
 		"{0: 3}" /* Vclock. */
 	), "re-enter candidate state");
 
+	/* Decrease election timeout to earlier than now. */
+
+	raft_run_for(election_timeout / 2);
+	raft_node_cfg_election_timeout(&node, election_timeout / 4);
+	ts = raft_time();
+	raft_run_next_event();
+
+	ok(raft_time() == ts, "the new timeout acts immediately");
+	ok(raft_node_check_full_state(&node,
+		RAFT_STATE_CANDIDATE /* State. */,
+		0 /* Leader. */,
+		5 /* Term. */,
+		1 /* Vote. */,
+		5 /* Volatile term. */,
+		1 /* Volatile vote. */,
+		"{0: 4}" /* Vclock. */
+	), "re-enter candidate state");
+
 	/*
 	 * Timeout smaller than a millisecond. Election random shift has
 	 * millisecond precision. When timeout is smaller, maximal shift is
diff --git a/test/unit/raft.result b/test/unit/raft.result
index 3fa2682c8..fcd180cfc 100644
--- a/test/unit/raft.result
+++ b/test/unit/raft.result
@@ -148,7 +148,7 @@ ok 7 - subtests
 ok 8 - subtests
 	*** raft_test_heartbeat: done ***
 	*** raft_test_election_timeout ***
-    1..11
+    1..13
     ok 1 - election is started
     ok 2 - enter candidate state
     ok 3 - new election is started
@@ -158,8 +158,10 @@ ok 8 - subtests
     ok 7 - new election timeout is respected
     ok 8 - but not too late
     ok 9 - re-enter candidate state
-    ok 10 - term is bumped, timeout was truly random
-    ok 11 - still candidate
+    ok 10 - the new timeout acts immediately
+    ok 11 - re-enter candidate state
+    ok 12 - term is bumped, timeout was truly random
+    ok 13 - still candidate
 ok 9 - subtests
 	*** raft_test_election_timeout: done ***
 	*** raft_test_election_quorum ***
-- 
2.24.3 (Apple Git-128)



More information about the Tarantool-patches mailing list